Update app.py

app.py

@@ -14,6 +14,7 @@ from sklearn.metrics.pairwise import cosine_similarity
 
 import networkx as nx
 from transformers import AutoTokenizer, TFAutoModelForSeq2SeqLM
+from summarizer import Summarizer,TransformerSummarizer
 
 nltk.download('punkt')
 nltk.download('stopwords')
@@ -34,6 +35,15 @@ model.load_weights("./marian_model/model.weights.h5")
 with open("cleaned_word_embeddings.pkl", "rb") as f:
     cleaned_word_embeddings = pickle.load(f)
 
+summ_model = TransformerSummarizer(transformer_type="XLNet",transformer_model_key="xlnet-base-cased")
+
+def translate_pretrained(text):
+    summarized = ''.join(summ_model(text))
+    tokenized = tokenizer([summarized], return_tensors="np")
+    out = model.generate(**tokenized)
+    arabic = tokenizer.decode(out[0], skip_special_tokens=True)
+    return arabic
+
 def get_clean_sentences(text):
     sentences = sent_tokenize(text)
     # Remove punctuations, numbers and special characters
@@ -116,6 +126,6 @@ def translate(text):
     arabic = tokenizer.decode(out[0], skip_special_tokens=True)
     return arabic
     
-demo = gr.Interface(fn=translate, inputs="text", outputs="text")
+demo = gr.Interface(fn=translate_pretrained, inputs="text", outputs="text")
 demo.launch(share=True)